Deep learning Training Institutes In Gurgaon Archives - DexLab Analytics | Big Data Hadoop SAS R Analytics Predictive Modeling & Excel VBA

Funnel Activation for Visual Recognition: A New Research Breakthrough

Funnel Activation for Visual Recognition: A New Research Breakthrough

The latest research work in the field of image recognition led to the development of a new activation function for visual recognition tasks, namely Funnel activation(FReLU). In this research ReLU and PReLU are extended to a 2D activation by adding a negligible overhead of spatial condition. Experiments on ImageNet, COCO detection, and semantic segmentation tasks are conducted to measure the performance of FReLU.

CNNs have shown advanced performances in many visual recognition tasks, such as image classification, object detection, and semantic segmentation.  In a CNN framework, basically two major kind of layers play crucial roles, the convolution layer and the non-linear activation layer. Both the convolution layers and activation layers perform distinct functions, however, in both layers there are challenges regarding capturing the spatial dependency. However, despite advancements achieved by complex convolutions, improving the performance of visual tasks is still challenging which results in Rectified Linear Unit (ReLU) being the most widely used function till date.

The research focused on two distinct queries

  1. Could regular convolutions achieve similar accuracy, to grasp the challenging complex images?
    2. Could we design an activation specifically for visual tasks?

1. Effectiveness and generalization performance

In a bid to find answers to these questions, researchers identified spatially insensitiveness in activations to be the main impending factor that prevent visual tasks from improving further.

To address this issue they proposed to find a new visual activation task that could be effective in removing this obstacle and be a better alternative to previous activation approaches.

How other activations work

Taking a look at other activations such as Scalar activations, Contextual conditional activations helps in understanding the context better.

Scalar activations basically are concerned with single input and output which could be represented in form of y = f(x). ReLU or, the Rectified Linear Unit is a widely used activation that is used for various tasks and could be represented as y = max(x, 0).

Contextual conditional activations work on the basis of many-to-one function. In this process neurons that are conditioned on contextual information are activated.

Spatial dependency modeling

In order to accumulate the various ranges of spatial dependences, some approaches utilize various shapes of convolution kernels which leads to lesser efficiency. In other methods like STN, spatial transformations are adaptively used for refining short-range dependencies for the dense vision tasks.

FReLU differs from all other methods in the sense that it performs better without involving complex convolutions. FReLU addresses the issues and solves with a higher level of efficiency.

Receptive field: How FReLU differs from other methods regarding the Receptive field

The size as well as the region of the receptive field play a crucial role in vision recognition tasks. The pixel contribution can be unequal. In order to implement the adaptive receptive field and for a better performance, many methods resort to complex convolutions. FReLU differs from such methods in the way that it achieves the same goal with regular convolutions in a more simple yet highly efficient manner.

Funnel Activation: how funnel activation works

FReLU being conceptually simple is designed for visual tasks. The research further delves into reviewing the ReLU activation and PReLU which is an advanced variant of ReLU, moving on to the key elements of FReLU the funnel condition and the pixel-wise modeling capacity, both of which are not found in ReLU or, in any of its variants.

2. Funnel activation

Funnel condition

Here the same max(·) is adopted as the simple non-linear function, when it comes to the condition part it gets extended to be a 2D condition which is dependent on the spatial context for individual pixel.  For the implementation of the spatial condition, Parametric Pooling Window is used for creating dependency.

Pixel-wise modeling capacity

 Due to the funnel condition the network is capable of generating spatial conditions in the non-linear activations for each pixel. This differs from usual methods where spatial dependency is created in the convolution layer and non-linear transformations are conducted separately. This model achieves a pixel-wise modeling capacity thereby extraction of spatial structure of objects could be addressed naturally.


Evaluation of the activation is tested via experiments on ImageNet 2012 classification dataset[9,37].The evaluation is done in stages starting with  different sizes of ResNet. Comparisons with scalar activations is done on ResNets with varying depths, followed by Comparison on light-weight CNNs. An object detection experiment is done to evaluate the generalization performance on various tasks on COCO dataset containing 80 object categories. Further comparison is also done on semantic segmentation task in CityScape dataset. Difference of the images could be perceived through the CityScape images.

Data Science Machine Learning Certification

4. Visualization of semantic segmentation

Funnel activation: ablation studies

The scope of the visual activation is tested further via ablation studies where each component of the activation namely 1) funnel condition, and 2)max(·) non-linearity are individually examined. The three parts of the investigation are as follows Ablation on the spatial condition, Ablation on the non-linearity, Ablation on the window size

Compatibility with Existing Methods

Before the new activation could be adopted into the convolutional networks, layers and stages need to be decided, the compatibility with other existing approaches such as SENet also was tested. The process took place in stages as follows

Compatibility with different convolution layers

Compatibility with different stages

Compatibility with SENet

Conclusion: Post all the investigations done to test out the compatibility of FReLU on different levels, it could be stated that this funnel activation is simple yet highly effective and specifically developed for visual tasks.  Its pixel-wise modeling capacity is able to grasp even complex layouts easily. But further research work could be done to expand its scope as it definitely has huge potential.

To get in-depth knowledge regarding the various stages of the research work on Funnel Activation for Visual Recognition, check



5 Most Powerful Computer Vision Techniques in use

5 Most Powerful Computer Vision Techniques in use

Computer Vision is one of the most revolutionary and advanced technologies that deep learning has birthed. It is the computer’s ability to classify and recognize objects in pictures and even videos like the human eye does. There are five main techniques of computer vision that we ought to know about for their amazing technological prowess and ability to ‘see’ and perceive surroundings like we do. Let us see what they are.

Image Classification

The main concern around image classification is categorization of images based on viewpoint variation, image deformation and occlusion, illumination and background clutter. Measuring the accuracy of the description of an image becomes a difficult task because of these factors. Researchers have come up with a novel way to solve the problem.

They use a data driven approach to classify the image. Instead of classifying what each image looks like in code, they feed the computer system with many image classes and then develop algorithms that look at these classes and “learn” about the visual appearance of each class. The most popular system used for image classification is Convolutional Neural Networks (CNNs).

Object Detection

Object detection is, simply put, defining objects within images by outputting bounding boxes and labels or tags for individual objects. This differs from image classification in that it is applied to several objects all at once rather than identifying just one dominant object in an image. Now applying CNNs to this technique will be computationally expensive.

So the technique used for object detection is region-based CNNs of R-CNNs. In this technique, first an image is scanned for objects using an algorithm that generates hundreds of region proposals. Then a CNN is run on each region proposal and only then is each object in each region proposal classified. It is like surveying and labelling the items in a warehouse of a store.

Object Tracking

Object tracking refers to the process of tracking or following a specific object like a car or a person in a given scene in videos. This technique is important for autonomous driving systems in self-driving cars. Object detection can be divided into two main categories – generative method and discriminative method.

The first method uses the generative model to describe the evident characteristics of objects. The second method is used to distinguish between object and background and foreground.

Semantic Segmentation

Crucial to computer vision is the process of segmentation wherein whole images are divided or segmented into pixelgroups that are subsequently labeled and classified.

The science tries to understand the role of each pixel in the image. So, for instance, besides recognizing and detecting a tree in an image, its boundaries are depicted as well. CNNs are best used for this technique.

Instance Segmentation

This method builds on semantic segmentation in that instead of classifying just one single dominant object in an image, it labels multiple images with different colours.

When we see complicated images with multiple overlapping objects and different backgrounds, we apply instance segmentation to it. This is done to generate pixel studies of each object, their boundaries and backdrops.

Data Science Machine Learning Certification


Besides these techniques to study and analyse and interpret images or a series of images, there are many more complex techniques that we have not delved into in this blog. However, for more on computer vision, you can peruse the DexLab Analytics website. DexLab Analytics is a premiere Deep Learning training institute In Delhi.



Using Deep Learning To Track Tropical Cyclones: A Study

Using Deep Learning To Track Tropical Cyclones: A Study

The severe cyclonic storm Nisarga approached the Maharashtra coast around Alibagh in Raigadh with “a sustained wind speed of 100-110 kmph” on June 3, 2020. Then it made landfall at Alibagh at around noontime. Landfall simply means that the storm, after having intensified over the ocean, has moved on to land.

Though the storm mellowed down in intensity as it approached the Maharashtra coast, government bodies took all precautions and evacuation work was done in advance on the basis of forecasts done by meteorologists and scientists.

To save lives and property, it is imperative to predict cyclones and the intensity with which they will strike. Deep Learning, a branch of artificial Intelligence, is helping scientists make breakthroughs in the science of forecasting cyclones.

Image Source:

Existing Storm Forecast Models

Most conventional dynamical models make accurate short term predictions but they are computationally demanding and “current statistical forecasting models have much room for improvement given that the database of past hurricanes is constantly growing”, says a report.

A tropical cyclone forecast involves the prediction of several interrelated features like track, intensity, rainfall, storm surge etc. The development of current hurricane and cyclone forecasts have advanced over the years but they are largely statistical in nature. The main limitation of this method is the complexity and non-linearity of atmospheric systems.

Deep Learning Models

Recurrent Neural Networks in deep learning models have been, of late, used to study increasingly complicated systems instead of the traditional methods of forecasting because they promise more accuracy. RNNs are a class of artificial neural networks where the modification of weights allows the model to learn intricate dynamic temporal behaviours, says another report.

An RNN with the capability of modelling complex non-linear temporal relationships of a hurricane or a cyclone could increase the accuracy of predicting future cyclonic path forecasts.

Machine Learning

Generally speaking, there are two methods or approaches to detecting extreme weather events like tropical cyclones – the data driven method which includes machine learning and the model driven approach which includes numerical simulation.

“The model-driven approach has the limitation that the prediction error increases with lead time because numerical models are inherently dependent on initial values. On the other hand, machine learning, as a data-driven approach, requires a large amount of high-quality training data,” says a report.

High quality data is easy to procure given the large amounts of data generated from weather stations on a daily basis the world over. So the machine learning method is easier to work and generate results from.

Data Science Machine Learning Certification


So what was difficult to do, that is find suitable metrics to study and detect the path of tropical cyclones earlier, has now become easier to do and scientists have been able to achieve accuracy in their predictions through the use of neural networks and artificial intelligence in general. For more on the subject, do read our blog here and here. Dexlab Analytics is a premier Deep Learning training institute in Delhi.



8 Applications of AI and Machine Learning in our Daily Lives

8 Applications of AI and Machine Learning in our Daily Lives

Artificial intelligence (AI) and machine learning are today thought to be one of the biggest innovations since the microchip. With the advancement of the science of neural networks, scientists are making extraordinary breakthroughs in machine learning through what is termed as deep learning. These sciences are making life easier and more streamlined for us in more ways than one. Here are a few examples.

1. Smart Gaming

Artificial Intelligence and Machine Learning are used in smart gaming techniques, especially in games that primarily require the use of mental abilities like chess. Google DeepMind’s AlphaGo learnt to play chess, and defeat champions like Lee Sedol (in 2016) by not only studying the moves of masters but by learning how to play the game by practising against itself innumerable times.

2. Automated Transportation

When we fly in an airplane, we experience automated transportation in the sense that a human pilot is only flying the plane for a couple of minutes during take-off and landing. The rest of the flight is maneuvered by a Flight Management System, a synchronization of GPS, motion sensors and computer systems that track flight position. Google Maps has already revolutionized local transport by studying coordinates from smart phones to determine how fast or slow a vehicle is moving and therefore how much traffic there is on a given road at any point of time.

3. Dangerous Jobs

AI technology powered robots are taking over dangerous jobs like bomb disposal and welding. In bomb disposal, today, robots need to be controlled by humans. But scientists believe there will soon come a time when these tasks would be completed by robots themselves. This technology has already saved hundreds of lives. In the field of welding, a hazardous job which entails working in high levels of noise and heat in a toxic environment, robots are helping weld with greater accuracy.

Data Science Machine Learning Certification

4. Environmental Protection

Machine Learning and artificial intelligence run on big data, large caches of data and mind boggling statistics generated by computer systems. When put to use in the field of environmental protection, these technologies could be used to extract actionable solutions to untenable problems like environmental degradation. For instance, “IBM’s Green Horizon Project studies and analyzes environmental data from thousands of sensors and sources to produce accurate, evolving weather and pollution forecasts.”

5. Robots as Friends

A company in Japan has invented what it calls a robot companion named Pepper who can understand and feel emotions and empathy. Introduced in 2014, Pepper went on sale in 2015 and all the 1000 units were sold off immediately. “The robot was programmed to read human emotions, develop its own, and help its human friends stay happy,” a report says. Robots could also assist the aged in becoming independent and take care of themselves, says a computer scientist at Washington State University.

6. Health Care

Hospitals across the world are mulling over the adoption of AI and ML to treat patients so there are reduced instances of hospital related accidents and spread of diseases like sepsis. AI’s predictive models are helping in the fight against genetic diseases and heart ailments. Also, Deep Learning models which “quickly provide real-time insights and…are helping healthcare professionals diagnose patients faster and more accurately, develop innovative new drugs and treatments, reduce medical and diagnostic errors, predict adverse reactions, and lower the costs of healthcare for providers and patients.”

7. Digital Media

Machine learning has revolutionized the entertainment industry and technology has already found buyers in streaming services such as Netflix, Amazon Prime, Spotify, and Google Play. “ML algorithms are…making use of the almost endless stream of data about consumers’ viewing habits, helping streaming services offer more useful recommendations.”

These technologies will assist with the production of media too. NLP (Natural Language Processing) algorithms help write and compose trending news stories, thus cutting on production time. Moreover, a new MIT-developed AI model named Shelley “helps users write horror stories through deep learning algorithms and a bank of user-generated fiction.”

8. Home Security and Smart Stores

AI-integrated cameras and alarm systems are taking the home security world by storm. The cutting-edge systems “use facial recognition software and machine learning to build a catalog of your home’s frequent visitors, allowing these systems to detect uninvited guests in an instant.” Brick and Mortar stores are likely to adopt facial recognition for payments by shoppers. Biometric capabilities are largely being adopted to enhance the shopping experience.

Key Takeaway

AI is no longer the domain of fiction. It’s our new reality and is it no surprise then that it is revolutionizing our lives. Deep learning training institutes and Machine Learning courses in India along with Artificial Intelligence courses in Delhi abound because India too is attempting to make the most of the AI revolution.


Google’s Deep Learning Tool Now Increases Accuracy for Breast Cancer Detection

Google’s Deep Learning Tool Now Increases Accuracy for Breast Cancer Detection

Google has finally developed a deep learning tool that identifies breast cancer that has spread to lymph nodes in pathology slides with 99% accuracy. It would surely reduce the average slide review time.

Detecting how far cancer has spread within a patient’s body is a Herculean task. Especially, for breast cancer. In this case, we’ve to detect how far cancer has spread from a primary region to neighboring lymph nodes. Nodal metastasis is the key here. It influences observations circulating radiation and chemotherapy, resulting in timely and proper detection.

Nevertheless, clinicians have always struggled to determine correctly how far the disease has spread. Fortunately, Google’s AI team proved better and productive at determining metastatic breast cancer with a greater accuracy. Two research papers by Google AI team have implemented deep learning methods to address the consequential challenge, and have lent a helping hand to the pathologists for effectively detecting breast cancer.


An algorithm, known as LYNA, Lymph Node Assistant has been developed to identify the regions of tumors that have spread or metastasized. Till now, they were extremely difficult to be detected by normal clinicians. As a well-known fact, out of half a million deaths across the globe owing to breast cancer, more than 90% are as a result of metastasis.

The abovementioned technology from Google first appeared in 2017. According to a recent publication, the AI research team at Google was influenced by “gigapixel-sized pathology slides of lymph nodes from breast cancer patients” for curating such an advanced algorithm. Moreover, the blog post revealed that the system was also able to “accurately pinpoint the location of both cancers and other suspicious regions within each slide.” In some cases, the locations are so minute that pathologists may have a hard time trying to detect them accurately.

The best part about LYNA system is regarding the area of concern for clinicians, doctors and how to enhance the entire process of review and ultimate diagnosis. According to Google, the underlying principle of this technology is to help doctors detecting metastatic breast cancer instead of replacing the human workforce. Thanks to the study and of course LYNA, the pathologists are in a better shape to accurately detect the micrometastases.

“Pathologists with LYNA assistance were more accurate than either unassisted pathologists or the LYNA algorithm itself,” reveals the blog post. This means the algorithms will become more productive when implemented by people, rather than working on their own.

However, the robust deep learning technology in question here does have some limitations – it works for limited dataset sizes. Further, only a single lymph node was scrutinized for every patient rather than multiple slides that would be common for a comprehensive clinical case. Thus, more detailed work needs to be done on LYNA before being applied to real-life patient situations.

For a detailed report, study “Artificial Intelligence Based Breast Cancer Nodal Metastasis Detection: Insights into the Black Box for Pathologists” as well as “Impact of Deep Learning Assistance on the Histopathologic Review of Lymph Nodes for Metastatic Breast Cancer.”

To know more about deep learning and how machine learning fuels the state of the art technology of deep learning, enroll in Deep Learning Training in Gurgaon. DexLab Analytics is one of the well-recognized deep learning training institutes in Delhi that offers in-demand skill training courses. For more information, visit their official site now.


The blog has been sourced from —


Interested in a career in Data Analyst?

To learn more about Data Analyst with Advanced excel course – Enrol Now.
To learn more about Data Analyst with R Course – Enrol Now.
To learn more about Big Data Course – Enrol Now.

To learn more about Machine Learning Using Python and Spark – Enrol Now.
To learn more about Data Analyst with SAS Course – Enrol Now.
To learn more about Data Analyst with Apache Spark Course – Enrol Now.
To learn more about Data Analyst with Market Risk Analytics and Modelling Course – Enrol Now.

Deep Learning: A Comprehensive Study

Deep Learning: A Comprehensive Study

Deep Learning is a subdivision of machine learning, under the category of artificial intelligence. It’s based on a fixed set of algorithms that strives to model advanced level abstractions in data. In a simple model, you would be having two sets of neurons, where if the input layer receives any input, it transmits a revamped version of input to the next layer. However, in a deep network, there exists a web of many layers between input and output, compelling the algorithm to rely on multiple processing layers, made of numerous layers and non-linear transformations.

No wonder, Deep Learning has triggered a revolution in the machine learning realm. Interesting works are being carried on in this field. Innovative technology is modifying speech recognition, object detection, visual object recognition and other sectors, like genomics and drug discovery. And, yes, we are excited about all the new good things that’s happening around!!

For more detailed analysis, scroll below:

About Deep Learning Architecture

  • Generative deep architectures are created to characterize high-order correlation attributes of visible data for all sorts of pattern analysis as well as synthetic purposes.
  • Discriminative deep architectures are specialized in offering discriminative power for pattern classification, mostly by showcasing posterior distribution of classes subject to visible data.
  • Hybrid deep architectures are designed for discrimination but are aided with results of generative architectures through better optimization as well as regularization.

A Few Applications of Deep Learning


Colorization of BW Images

Deep learning has the ability to recreate an image with the addition of color. The cutting edge technology uses the objects and the entire context within a picture for coloring the whole image, quite similar to a human approach. For this, extensive supervised layers and convulational neural network have to be put to use, of course.

Generative Model Chatbots

They are in hype. A sequence-to-sequence model is widely used to design chatbots which are capable of generating their own answer when trained on a wide set of real-live interactive datasets.

Machine Translations

Text translation is very easy to perform without following any proper sequence, allowing algorithms to ace dependencies between words and plotting to a new language.

Automatic Game Playing

Here, a model is trained to play a computer game formulated on the pixels on the screen. The task is fairly challenging and is one of the most fascinating domains of deep reinforcement models, Deep Mind.

Automatic Handwriting Generation

Here, you have to generate a new handwriting for a particular word or phrase using this technology. The handwritting is given as a sequence of coordinates written by a pen once the samples are done.

As parting thoughts, Deep Learning is still in a nascent stage in India. But, its diverse uses and capabilities will surely put it in the industry frontline some day soon. So, if you are looking for good deep learning training courses in Gurgaon, DexLab Analytics offers some out of the box kind of learning experience. Do check out their deep learning certification courses, they are excellent!


The blog has been sourced from —

Interested in a career in Data Analyst?

To learn more about Data Analyst with Advanced excel course – Enrol Now.
To learn more about Data Analyst with R Course – Enrol Now.
To learn more about Big Data Course – Enrol Now.

To learn more about Machine Learning Using Python and Spark – Enrol Now.
To learn more about Data Analyst with SAS Course – Enrol Now.
To learn more about Data Analyst with Apache Spark Course – Enrol Now.
To learn more about Data Analyst with Market Risk Analytics and Modelling Course – Enrol Now.

Call us to know more